Fast Intra-prediction Mode Decision Of H.264 And Transform Algorithms Research

H.264,a video coding standard developed by the Joint Video Team (JVT) of International Telecommunication Union (ITU) and International Standards Organization (ISO), is the latest standard for the practical application. The goal is to achieve high video resolution, improve image quality and be able to cover both low bandwidth and high bandwidth application. The compression performance and structural design of H.264 are much better than the past standards. More than that, it has unparalleled advantages such as low code rate, high image quality, strong fault tolerance, fine network adaptability and so on. However, H.264 is based on the cost of increased complexity, which makes it difficult to apply in real-time video processing.Thus, various optimized algorithms for techniques in H.264 emerged. Due to a variety of forecasting modes used in intra prediction and the optimization of rate distortion process's great influence on the performance of the encoder, H.264 has a high computational complexity, resulting in the slow encoding rate. This paper analyzes the H.264/AVC intra prediction mode decision process and proposes a fast intra₄×4 mode decision algorithm. This algorithm is based on the relevance of adjacent block prediction modes and the relevance of the various prediction modes, does the pre-judgment among 9 modes before computing the RD_cost, chooses the modes related to the most likely mode as candidate modes to reduce the complexity of intra prediction while basically maintaining the coding performance of H.264.In image coding, transform coding and quantization are two separate processes in principle. In H.264, the two processes merge into one process and use integer operations, reducing the encoding and decoding operations, improving real-time image compression. Quantization process determines the quantitative parameters according to the size of the image dynamic range, which not only retains the necessary details of the image, but also reduces the code stream.Block-based hybrid coding is the basic coding scheme for H.261,H.263,H.264,JPEG and MPEG, but it leads to noticeable block effect under coarse quantization factors. This paper proposes an reducing the block effect in both the frequency domain and spatial domain. For smooth regions, our method take advantage of the fact that the continuity of the original pixels and the correlation of adjacent blocks to reduce the cross-border discontinuity of the pixels. For the edge regions, a smoothing filter is used to retain the edge. Simulation results show that this method can preserve the main structure of image, smoothing noise and removing eliminate the block effects and implement low video code rate.